| Age | Commit message (Collapse) | Author |
|---|
|
Add KUnit tests for hdmi_frl_status_polling_work() covering the
no-links, no-local-sink, non-HDMI and zero-FRL-rate skip cases, plus
the poll paths that leave the link unchanged or trigger a retrain.
Assisted-by: Copilot:Claude-Opus-4.8
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
[Why&How]
The KUnit connector test references hdmi_frl_status_polling_work(), but the
function was static with no visible prototype, causing a build failure
(undeclared / implicit declaration). Make the function non-static and add
a prototype to amdgpu_dm_connector.h so the test can reference it. The
INIT_DELAYED_WORK setup remains in amdgpu_dm_connector_init_helper(); no
runtime behavior change.
Assisted-by: Copilot:Claude-Opus-4.8
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
Add a KUnit test for add_fs_modes() confirming that a generated mode
with illegal timing is skipped rather than added.
Assisted-by: Copilot:Claude-Opus-4.8
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
Add KUnit tests for add_fs_modes() covering successful freesync-video
mode generation and refresh rates that fall outside the supported
range.
Assisted-by: Copilot:Claude-Opus-4.8
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
Add KUnit tests for amdgpu_dm_prune_primary_tile_modes() covering the
no-sink, no panel-patch and non-tiled early returns, a secondary tile,
and the primary-tile path that prunes per-tile modes.
Assisted-by: Copilot:Claude-Opus-4.8
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
[WHY&HOW]
For cases where number of OPP/DPP mismatch, like 4 planes, 1 stream on
a 3 OPP ASIC, mode support would previously fail citing lack of OPP's.
This is incorrect, as the configuration only needs 1 OPP but 4 DPP's.
Reviewed-by: Austin Zheng <austin.zheng@amd.com>
Signed-off-by: Dillon Varone <Dillon.Varone@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
[WHAT]
Add KUnit tests for amdgpu_dm_dcn10_register_irq_handlers() covering the
add_id() rejections for the VUPDATE and HPD sources and the
invalid-source guard in the VUPDATE_NO_LOCK loop, and for
amdgpu_dm_register_outbox_irq_handlers() rejecting a source with no
funcs table.
[HOW]
Reuse the shared registration fixture and the fake IRQ service that maps
no source IDs.
Assisted-by: Copilot:Claude-Opus-5
Reviewed-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
[WHAT]
Add KUnit tests for amdgpu_dm_dce110_register_irq_handlers() covering
the SOC15 DCE client used from Vega on, the amdgpu_irq_add_id()
rejections for the CRTC, VUPDATE and HPD sources, and the
invalid-source guards in the VBLANK, VUPDATE and PFLIP loops.
[HOW]
Add three fake IRQ services that map only a prefix of the source IDs so
each loop reaches its guard in turn. Clearing a single amdgpu_irq_src's
funcs makes only that source's add_id() fail.
Assisted-by: Copilot:Claude-Opus-5
Reviewed-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
[WHAT]
Add KUnit tests for dm_dmub_outbox1_low_irq() draining a trace ring up
to the DMUB_TRACE_MAX_READ cap, dm_handle_hpd_work() rejecting a work
item with no notification, and amdgpu_dm_irq_schedule_work() not
queueing once the IRQ workqueue is gone.
[HOW]
Back dmub->outbox0_rb with a fake ring whose write pointer sits one
entry past the cap, and keep the ring larger so the read pointer does
not wrap. dm_handle_hpd_work() returns before its own kfree() on the
NULL-notify path, so the test frees the work item.
Assisted-by: Copilot:Claude-Opus-5
Reviewed-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
[WHAT]
Add KUnit tests for handle_hpd_rx_irq() re-detecting a downstream port
change on a non-MST-root connector, amdgpu_dm_hpd_init() skipping a
connector with no dc_link, and dmub_hpd_callback() deferring HPD while
suspended. Also add a writeback connector to the HPD registration test.
[HOW]
The downstream-change test reuses dm_test_setup_hpd_rx_irq() and clears
the MST root state, with dc_link_detect() stubbed to fail. The suspend
test leaves adev->dm.ddev NULL, so reaching the connector walk would
crash.
Assisted-by: Copilot:Claude-Opus-5
Reviewed-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
[WHAT]
Add KUnit tests for handle_hpd_irq_helper() covering a failing
dc_link_detect_connection_type() and the early return taken when a
debounce re-detect is already scheduled. Also arm the debounce work in
the stale-prev-sink test so mod_delayed_work() re-schedules.
[HOW]
Both build on dm_test_setup_hpd_irq_helper() and override only the
link_srv stubs they need. The pending-debounce test arms the work ten
seconds out and spies on dc_link_detect() to prove it was skipped.
Assisted-by: Copilot:Claude-Opus-5
Reviewed-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
[WHAT]
Add KUnit tests for schedule_dc_vmin_vmax(), which had no coverage. Both
callers are exercised: dm_crtc_high_irq() on an AI+ family and the DCE
path of dm_vupdate_high_irq() on a pre-DCE12 family.
[HOW]
Add a dm_test_setup_vmin_vmax_crtc() fixture building a CRTC with an
active VRR stream and a vmin/vmax work queue. Each test flushes the
queue and checks the worker applied the adjust and dropped its stream
reference.
Assisted-by: Copilot:Claude-Opus-5
Reviewed-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
[WHAT]
Add KUnit tests for the two writeback vblanks in dm_crtc_high_irq(): the
first disables frame capture and marks wb_frame_done, the second signals
the out fence and clears the pending state.
[HOW]
dc_stream_fc_disable_writeback() dereferences dc->res_pool before its
NULL-stream check, so supply a dc with a fake IRQ service. The
completion test queues a bare drm_writeback_job and holds a vblank
reference for the completion's drm_crtc_vblank_put().
Assisted-by: Copilot:Claude-Opus-5
Reviewed-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
[WHAT]
Add KUnit tests for pageflip completion: the non-VRR path in
dm_pflip_high_irq(), and the DCN VUPDATE_NO_LOCK paths in
dm_vupdate_high_irq() that deliver the event once the OTG reports no
flip pending, or retire the flip when there are no active planes.
[HOW]
Initialise DRM vblank and hold a reference to balance the handler's
drm_crtc_vblank_put(). A non-zero DCE IP version selects the DCN path,
and a NULL adev->dm.dc makes dc_get_flip_pending_on_otg() report no
pending flip.
Assisted-by: Copilot:Claude-Opus-5
Reviewed-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
[WHAT]
Several amdgpu_dm_irq tests were each open-coding the same setup. Share
it through three functions: dm_test_add_crtc(),
dm_test_setup_hpd_irq_helper() and dm_test_setup_irq_regs().
Assisted-by: Copilot:Claude-Opus-5
Reviewed-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
Add KUnit tests for amdgpu_set_panel_orientation() covering the non-eDP
early return, the missing native mode case, and a panel that applies
the orientation quirk.
Assisted-by: Copilot:Claude-Opus-4.8
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
Add KUnit tests driving create_eml_sink() through the valid-EDID
branch: reading the EDID over DDC, and the forced-on paths using an
emulated sink and a cached local sink.
Assisted-by: Copilot:Claude-Opus-4.8
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
Add KUnit tests for amdgpu_dm_connector_get_modes() covering the
no-EDID default modes, the extra modes added for 128b/132b and analog
links, and the path that adds modes from a valid EDID.
Assisted-by: Copilot:Claude-Opus-4.8
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
Add a KUnit test that drives amdgpu_dm_connector_funcs_force() through
the valid-EDID branch: a fake i2c adapter serves a valid EDID over the
DDC read protocol so the connector caches drm_edid and, with an emulated
sink and dc_link present, copies the raw EDID into the sink and parses
its caps.
Assisted-by: Copilot:Claude-Opus-4.8
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
Add KUnit tests for the amdgpu_dm_connector_atomic_check() properties
that force a modeset: privacy-screen, colorspace and content-type
changes, plus the HDR-metadata exit and fill-error branches.
Assisted-by: Copilot:Claude-Opus-4.8
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
Add KUnit tests for amdgpu_dm_connector_atomic_check() covering the
no-CRTC early return, the DP-MST connector path, and an unchanged state
that skips revalidation.
Assisted-by: Copilot:Claude-Opus-4.8
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
Add KUnit tests for amdgpu_dm_fill_hdr_info_packet() covering the HDMI
and DisplayPort HDR paths, an unsupported connector signal, and invalid
HDR metadata.
Assisted-by: Copilot:Claude-Opus-4.8
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
Add a KUnit test driving amdgpu_dm_connector_mode_valid() through the
EDID-management branch for a forced-on connector.
Assisted-by: Copilot:Claude-Opus-4.8
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
[Why&How]
MCM and RMCM both took their cube from struct tetrahedral_params, which is
embedded by value in dc_3dlut and therefore in dc_plane_state and dc.scratch.
RMCM needs a 33x33x33 cube, but that cube is ~421KB, so enabling it in the
shared union - where tetrahedral_33 was left commented out - would push
sizeof(struct dc) past the kmalloc MAX_PAGE_ORDER limit and break kunit
and other allocations.
Drop the unused struct tetrahedral_33x33x33 definition and lut_dim.
Reviewed-by: Dillon Varone <dillon.varone@amd.com>
Reviewed-by: Sridevi Arvindekar <sridevi.arvindekar@amd.com>
Signed-off-by: Rafal Ostrowski <rafal.ostrowski@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
[WHAT]
Add KUnit tests for amdgpu_dm_plane_handle_cursor_update() and
amdgpu_dm_plane_atomic_async_update() covering the return taken when the
cursor plane has no framebuffer, and the copy of the new position and
size into the plane state ahead of the cursor update.
Assisted-by: Copilot:Claude-Opus-5
Reviewed-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
[WHAT]
Add KUnit tests for amdgpu_dm_plane_helper_prepare_fb() and
amdgpu_dm_plane_helper_cleanup_fb() covering the early return taken when
the plane state carries no framebuffer, so no buffer object is pinned or
unpinned.
Assisted-by: Copilot:Claude-Opus-5
Reviewed-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
[WHAT]
Add KUnit tests for amdgpu_dm_plane_get_plane_modifiers() and
amdgpu_dm_plane_format_mod_supported() on a GFX6-8 device. The exposed
list holds one macro tiled modifier per distinct bits per pixel plus a
micro tiled and a linear one, and a listed modifier is validated through
the GFX6-8 specific check.
Add dm_test_mods_count() for the list length and use it in the existing
sentinel scans.
Assisted-by: Copilot:Claude-Opus-5
Reviewed-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
[WHAT]
Add KUnit tests for amdgpu_dm_plane_calc_gfx6_mod() and
amdgpu_dm_plane_gfx6_format_mod_supported() covering the micro tiled
early return, the GFX7-8 macro tile mode table lookup, the GFX6 tile
mode register fields, and the rejection of foreign, DCC, newer tile
version, GFX9+ and non-displayable modifiers.
Assisted-by: Copilot:Claude-Opus-5
Reviewed-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
[WHAT]
Add KUnit tests for amdgpu_dm_plane_get_gfx6_tile_idx(),
amdgpu_dm_plane_calc_gfx7_tile_split() and
amdgpu_dm_plane_get_gfx7_macro_tile_idx() covering the per-family tile
mode index selection, the tile split derived from the sample split
factor with its 256 byte and memory row size clamps, and the macro tile
index capped by the tile split.
Assisted-by: Copilot:Claude-Opus-5
Reviewed-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
[WHAT]
Add KUnit tests for amdgpu_dm_plane_fill_gfx6_tiling_info_from_modifier()
covering the linear, micro tiled and macro tiled modifiers, and the
rejection of a modifier that is not of the GFX6 tile version.
Assisted-by: Copilot:Claude-Opus-5
Reviewed-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
[WHAT]
Add KUnit tests for amdgpu_dm_plane_add_modifier_dedup() covering a
modifier already present in the list, which is not appended a second
time, a new modifier, which is appended, and a NULL list left over from
an earlier allocation failure, which is a no-op.
Assisted-by: Copilot:Claude-Opus-5
Reviewed-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
[WHAT]
Add KUnit tests for amdgpu_dm_plane_drm_plane_duplicate_state() and
amdgpu_dm_plane_drm_plane_destroy_state() covering the resource paths.
Duplicating a state shares every color blob and takes an extra reference
on the DC plane state, and destroying one drops a reference on each.
Assisted-by: Copilot:Claude-Opus-5
Reviewed-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
[WHAT]
Add KUnit tests for dm_dp_mst_connector_destroy() covering the teardown
of a connector that never had a remote sink, and the teardown that first
removes the remote sink from the DC link.
[HOW]
The function frees both the connector and the MST port, so the test
allocates them outside the KUnit managed allocator. The parent branch
device takes an extra malloc reference so it survives the port teardown.
Assisted-by: Copilot:Claude-Opus-5
Reviewed-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
[WHAT]
Add KUnit tests for amdgpu_dm_mst_connector_late_register() and
amdgpu_dm_mst_connector_early_unregister(), covering the remote AUX bus
registration, the unregister that only resets the MST status when no
remote sink is attached, and the unregister that removes the sink from
the DC link and clears the MST connector state.
[HOW]
The tests reuse the MST downstream connector fixture, which already
provides the kernel device the DRM MST helpers log through and the
modeset lock the unregister path takes.
Assisted-by: Copilot:Claude-Opus-5
Reviewed-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
[WHAT]
Add KUnit tests for dm_dp_mst_detect() covering the DPCD revision probe
of a peer device, the fallback when both revision registers read back as
zero, the NAKed remote read that must leave the cached revision alone,
and the unplugged port whose disconnected status releases the remote sink
and resets the MST connector state.
[HOW]
The tests reuse the MST downstream connector fixture. Its empty topology
manager makes drm_dp_mst_detect_port() report a disconnected port, which
is what the unplug path needs, and the port AUX is served by the fake
DPCD store.
Assisted-by: Copilot:Claude-Opus-5
Reviewed-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
[WHAT]
Add KUnit tests for dm_dp_mst_get_modes() covering the paths taken once
an EDID is available: a placeholder virtual sink is released and replaced
by a real remote sink, DC refusing that sink returns zero modes, the
content protection state is restored from the HDCP workqueue, and an EDID
read from a port still in the topology is cached with MST_REMOTE_EDID
flagged.
[HOW]
The tests reuse the MST downstream connector fixture and feed it a
minimal EDID base block. For the remote read the fixture topology gains a
branch device holding the port so drm_dp_mst_edid_read() can validate it
and return the cached EDID.
Assisted-by: Copilot:Claude-Opus-5
Reviewed-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
[WHAT]
Add KUnit tests for dm_dp_mst_get_modes() covering the path taken when
the remote EDID cannot be read: the default remote sink is registered and
MST_REMOTE_EDID is cleared, DC refusing the sink leaves the connector
without one, and an already attached sink is kept instead of asking DC
for a second one.
[HOW]
A shared fixture builds an MST downstream connector on a real DRM device
with an empty topology manager, so drm_dp_mst_edid_read() takes its "port
is gone" path without needing a branch device. dc_link_add_remote_sink()
and dc_link_remove_remote_sink() are thin wrappers over link_service
callbacks, so fakes for those two are enough to run the connector paths
without a DC core.
Assisted-by: Copilot:Claude-Opus-5
Reviewed-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
[WHAT]
Add KUnit tests for dm_handle_mst_sideband_msg_ready_event() covering the
DOWN_REP ready bit that is acknowledged at DPCD and re-polled until the
iteration limit, and the case where the acknowledge write keeps failing
and the handler gives up after the third retry.
[HOW]
A live topology manager is needed for the handler to reach its ack path,
so the tests initialise one on a DRM device and arm the DOWN_REP ready
bit in the fake DPCD. The fake AUX channel gains a separate write result
so the ESI read still succeeds while the acknowledge write fails.
Assisted-by: Copilot:Claude-Opus-5
Reviewed-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
Add KUnit tests for amdgpu_dm_connector_mode_valid() covering the
missing dc_sink path and a forced-on connector that fails to create a
validation stream.
Assisted-by: Copilot:Claude-Opus-4.8
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
Add KUnit tests for amdgpu_dm_create_validate_stream_for_sink()
covering the writeback connector path, the colour-depth downgrade when
no valid bpc is found, the HDMI YCbCr and forced YCbCr420/422/444
output paths, and the timing/context prune-and-retry loops.
Assisted-by: Copilot:Claude-Opus-4.8
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
Add dm_test_validate_stream_dc_ok_no_pipe which builds a minimal fake
dc so stream and plane validation both pass, then lets
dc_state_add_stream() bail out early via timing_generator_count == 0 to
avoid deep pipe allocation, raising coverage from 24% to 90%.
Assisted-by: Copilot:Claude-Opus-4.8
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
Add a KUnit test that drives amdgpu_dm_connector_detect() through the
amdgpu_dm_hide_secondary_tile_from_userspace() path using a sink that
requests disable_second_tile and a synthesized DisplayID tiled EDID at a
non-origin tile location, so the secondary Apple Studio Display tile is
hidden and the connector reports disconnected. Raises the helper coverage
from 44% to 100%.
Assisted-by: Copilot:Claude-Opus-4.8
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
Add tests covering the non-shortcut poll paths: connected reusing a cached
local_sink, connected via full dc_link_detect(), and the disconnect
teardown that releases and clears a stale sink. Fakes wire
link->dc->link_srv detect_connection_type/detect_link function pointers.
Raises amdgpu_dm_connector_poll coverage from 35% to 100%.
Assisted-by: Copilot:Claude-Opus-4.8
Reviewed-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
[Why]
Dynamic ODM lowers the required dispclk and saves power. DCN42
and the DCN42 variant should follow whatever the driver requests
through the config flag rather than defaulting it off in resource
construct.
[How]
Assign minimize_dispclk_using_odm directly from the config flag
so an enable request from the driver is honoured. No functional
change when the flag is not set.
Reviewed-by: Dmytro Laktyushkin <dmytro.laktyushkin@amd.com>
Signed-off-by: Gabe Teeger <gabe.teeger@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
[Description]
- Align calculation with addrlib (need to align to macro tiled
widht/height)
- Align the copy size / addr to 256KB for SW_256KB modes
Reviewed-by: Dillon Varone <dillon.varone@amd.com>
Signed-off-by: Alvin Lee <Alvin.Lee2@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
[Why&How]
DML should have a method of disabling alt-ch pstate method.
Add a debug mask bit to disable the alt-ch pstate method in PMO.
Reviewed-by: Dillon Varone <dillon.varone@amd.com>
Signed-off-by: Alvin Lee <Alvin.Lee2@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
[WHAT]
Add KUnit tests for dm_dmub_sw_init(), covering the ASIC to DMUB service
mapping for all 19 supported IP versions, the memory domain choice, a
complete successful init, the BSS data region variant and the PSP
firmware load registration.
[HOW]
The framebuffer allocation needs a live TTM device, so add
amdgpu_bo_create_kernel() to the indirection table introduced for the
firmware request. The ASIC mapping and memory domain tests force that
allocation to fail, which walks every ASIC through DMUB service creation
and region calculation without a fake framebuffer.
Grow the fake firmware so its instruction constant region is larger than
two PSP footers. The footer probing in
dmub_srv_get_fw_meta_info_from_raw_fw() walks back from the end of that
region twice, so a region smaller than the footers underflows the size
and reads out of bounds.
Assisted-by: Copilot:Claude-Opus-5
Reviewed-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
[WHAT]
Add KUnit tests for dm_init_microcode(), covering the ASIC to firmware
name mapping for all 21 supported IP versions and the firmware request
failure path.
[HOW]
The real amdgpu_ucode_request() reaches the firmware loader, so add a
KUnit-only indirection table for it. The default table entry points at
the existing function, so no test-only wrapper is introduced, and the
prototype is mirrored verbatim so an upstream change fails the build at
the default initializer rather than diverging silently.
Tests install a fake that records the formatted firmware name, letting a
single table-driven case assert the name chosen for every ASIC,
including the Green Sardine revision check and the Sienna Cichlid versus
Navy Flounder split on the graphics IP version.
Assisted-by: Copilot:Claude-Opus-5
Reviewed-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
[WHAT]
Add KUnit tests for dm_dmub_get_vbios_bounding_box(), covering the
successful handshake on every ASIC that has a bounding box, the GPU
memory allocation failure path, and both GPINT timeout paths. This also
covers dm_dmub_send_vbios_gpint_command().
[HOW]
Extend the fake CGS device with a fail_on_write hook so a chosen GPINT
write is never acked, which drives the timeout paths without waiting on
real firmware. Reuse the buffer object indirection added for
dm_allocate_gpu_mem() with a fake allocator, since the real path needs a
live TTM device.
The success test walks every ASIC with a bounding box and confirms a
non-zero GART buffer is allocated and that all four address words plus
the copy request are sent. The timeout tests confirm the buffer is
released and the allocation list left empty.
Assisted-by: Copilot:Claude-Opus-5
Reviewed-by: Bhawanpreet Lakha <bhawanpreet.lakha@amd.com>
Signed-off-by: Alex Hung <alex.hung@amd.com>
Signed-off-by: Ivan Lipski <ivan.lipski@amd.com>
Tested-by: Dan Wheeler <daniel.wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
|
A reset on init lowers the init level to the minimal XGMI set, which
leaves out the RAS IP block, so its hw_init never runs and the PSP
buffers stay unallocated. The resume that follows the reset tries to
reload the RAS firmware into those buffers, fails with -ENOMEM and
aborts the whole resume, taking every device of the hive down with it.
There is nothing to reload in that case. Leave the block alone and let
amdgpu_ras_resume_after_reset() bring it up, which is where it has to
happen anyway: the RAS hardware init reads the EEPROM and so takes the
reset domain lock for read, while the reset on init path still holds it
for write, so doing it here deadlocks.
Signed-off-by: Xiang Liu <xiang.liu@amd.com>
Reviewed-by: Hawking Zhang <Hawking.Zhang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
|
